def lung_slic_tool(filename): prename, extname = os.path.splitext(filename) pathname = os.path.split(filename)[0] if (extname == '.npy'): segresult = np.load(filename) labels = np.load(prename + '.label.npy') elif (extname == '.mhd'): npyname = filename + '.npy' filelist = [pathname + '/' + i for i in os.listdir(pathname)] if npyname in filelist: segresult = np.load(npyname) labels = np.load(filename + '.label.npy') else: full_image_info = sitk.ReadImage(filename) full_scan = sitk.GetArrayFromImage(full_image_info) old_spacing = np.array(full_image_info.GetSpacing())[::-1] image, new_spacing = mt.resample(full_scan, old_spacing) ##### #image = image[195:200] ##### print("slic segmenting") labels = lc.slic_segment(image, view_result=True) print("segmentation complete")
print('Detection Done. time:{}s'.format(time.time() - start_time)) if 'vision_path' in dir() and 'vision_path' is not None: nodules = [] for nc in range(len(nodule_center_predictions)): nodules.append(np.int_(nodule_center_predictions[nc][0:3])) volume_predicted = cvm.view_coordinations(result_predictions * 1000, nodules, window_size=10, reverse=False, slicewise=False, show=False) np.save(vision_path + "/" + uid + "_prediction.npy", volume_predicted) prediction_cluster_vision = lc.segment_color_vision( prediction_labels) np.save(vision_path + "/" + uid + "_prediction_clusters.npy", prediction_cluster_vision) ''' #randomly create a result for testing nodule_center_predictions = [] for nc in range(10): nodule_center_predictions.append([random.randint(0,image.shape[0]-1), random.randint(0,image.shape[1]-1), random.randint(0,image.shape[2]-1), random.random()]) ''' print('Nodule coordinations:') if len(nodule_center_predictions) <= 0: print('none') for nc in range(len(nodule_center_predictions)): #the output coordination order is [x,y,z], while the order for volume image should be [z,y,x] results.append([ uid, (nodule_center_predictions[nc][2] * new_spacing[2]) +
#mbb.z_top = 195 #mbb.z_bottom = 205 #mbb.y_top = 110 #mbb.y_bottom = 120 #mbb.x_top = 105 #mbb.x_bottom = 115 image_bounded = image[mbb.z_top:mbb.z_bottom+1, mbb.y_top:mbb.y_bottom+1, mbb.x_top:mbb.x_bottom+1] print('Lung Segmentation Done. time:{}s' .format(time.time()-start_time)) #nodule_matrix, cindex = cd.candidate_detection(segimage) #cluster_labels = lc.seed_volume_cluster(nodule_matrix, cluster_size=30, result_vision=False) num_segments = int(image_bounded.shape[0] * image_bounded.shape[1] * image_bounded.shape[2] / 27) #the volume of a 3mm nodule is 27 voxels #num_segments = 20 print('cluster number:%d' %(num_segments)) cluster_labels = 0 - np.ones(shape=image.shape, dtype=int) cluster_labels_bounded = lc.slic_segment(image_bounded, compactness=0.01, num_segments=num_segments) print('Clustering Done. time:{}s' .format(time.time()-start_time)) cluster_labels[mbb.z_top:mbb.z_bottom+1, mbb.y_top:mbb.y_bottom+1, mbb.x_top:mbb.x_bottom+1] = cluster_labels_bounded cluster_labels[np.logical_or(segmask==0, image<-600)] = -1 #cluster_labels_filtered_fast = lc.cluster_filter_fast(image, cluster_labels) #print('Cluster Filtering Fast Done. time:{}s'.format(time.time() - start_time)) #cluster_labels_filtered = lc.cluster_filter(image, cluster_labels) #the clusters with no tissue are filetered out #candidate_coords_slow, candidate_labels_slow = lc.cluster_centers_fast(cluster_labels_filtered) #volume_candidated = cvm.view_coordinations(image, candidate_coords_slow, window_size=10, reverse=False, slicewise=True, show=False) #np.save(vision_path+"/"+uid+"_candidate_slow.npy", volume_candidated) #print('candidate slow number:%d' %(len(candidate_coords_slow))) #print('Cluster Filtering Done. time:{}s'.format(time.time() - start_time)) #segresultfast = lc.segment_vision(image, cluster_labels_filtered_fast) #np.save(vision_path+"/"+uid+"_segresultfast.npy", segresultfast) if RESULT_VISION:
def annotations_crop(self, overbound=True, candsample=False): if os.access(self.output_path, os.F_OK): shutil.rmtree(self.output_path) os.makedirs(self.output_path) os.mkdir(self.nodules_npy_path) os.mkdir(self.nonnodule_npy_path) if not os.access(self.vision_path, os.F_OK): os.makedirs(self.vision_path) for patient in enumerate(tqdm(self.ls_all_patients[814:])): patient = patient[1] #patient = "./SLH_data/0721/285 0800418645" print(patient) full_scan, full_image_info, patient_uid = mt.read_dicom_scan( patient) if full_scan.min() < -1024: errorlog = open("results/error.log", "w") errorlog.write("Hu unit incorrect:%s\n" % (patient)) errorlog.close() origin = np.array( full_image_info.GetOrigin())[::-1] #---获取“体素空间”中结节中心的坐标 old_spacing = np.array(full_image_info.GetSpacing() )[::-1] #---该CT在“世界空间”中各个方向上相邻单位的体素的间距 min_space = old_spacing.min() image, new_spacing = mt.resample(full_scan, old_spacing) #---重采样 print('resample done') silist = self.df_annotations.serie_id.tolist() if silist.count(patient_uid) == 0: print('no annotation for this patient found') continue serie_index = silist.index(patient_uid) patient_nodules = [] for annocol in self.annotation_columns: annostr = self.df_annotations.get(annocol)[serie_index] if type(annostr) == unicode: #annotation = np.array(annostr.split(u'\uff08')[0].split(' '), dtype=int) #patient_nodules.append([serie_index, annotation]) #the index order is [x,y,z] if annostr.find(u'*') >= 0: continue coordbegin = -1 coordend = -1 for ci in range(len(annostr)): if coordbegin < 0: if annostr[ci] >= u'0' and annostr[ci] <= u'9': coordbegin = ci elif (annostr[ci] < u'0' or annostr[ci] > u'9') and annostr[ci] != u' ': coordend = ci break if coordbegin >= 0: if coordend < 0: coordend = len(annostr) coordstr = annostr[coordbegin:coordend] annotation = np.array(coordstr.split(u' '), dtype=int) patient_nodules.append( [annocol, annotation]) # the index order is [x,y,z] if type(annostr) == str: if annostr.find('*') >= 0: continue coordbegin = -1 coordend = -1 for ci in range(len(annostr)): if coordbegin < 0: if annostr[ci] >= '0' and annostr[ci] <= '9': coordbegin = ci elif (annostr[ci] < '0' or annostr[ci] > '9') and annostr[ci] != ' ': coordend = ci break if coordbegin >= 0: # annotation = np.array(annostr.split('(')[0].split(' '), dtype=int) if coordend < 0: coordend = len(annostr) coordstr = annostr[coordbegin:coordend] annotation = np.array(coordstr.split(' '), dtype=int) patient_nodules.append( [annocol, annotation]) # the index order is [x,y,z] v_centers = [] center_coords = [] for annocol, nodule in patient_nodules: nodule_center = np.array(np.flip(nodule, axis=0) * old_spacing / new_spacing, dtype=int) #---获取“世界空间”中结节中心的坐标 #v_center = np.rint((nodule_center - origin) / new_spacing) #映射到“体素空间”中的坐标 #v_center = np.array(v_center, dtype=int) v_centers.append([annocol, nodule_center]) center_coords.append(nodule_center) #volume_regioned = cv.view_coordinations(image, center_coords, window_size=10, reverse=False, slicewise=False, show=False) #cv.view_CT(volume_regioned) #np.save(self.vision_path+"/"+patient_uid+"_annotated.mhd", volume_regioned) #---这一系列的if语句是根据“判断一个结节的癌性与否需要结合该结节周边位置的阴影和位置信息”而来,故每个结节都获取了比该结节尺寸略大的3D体素 #get annotations nodule window_half = int(BOX_SIZE / 2) if overbound: box_size = 2 * BOX_SIZE box_half = BOX_SIZE else: box_size = BOX_SIZE box_half = window_half for annocol, v_center in v_centers: zyx_1 = v_center - box_half # 注意是: Z, Y, X zyx_2 = v_center + box_half if mt.coord_overflow(zyx_1, image.shape) or mt.coord_overflow( zyx_2, image.shape): zyx_1_fix = zyx_1.copy() zyx_2_fix = zyx_2.copy() for ci in range(3): if zyx_1[ci] < 0: zyx_1_fix[ci] = 0 elif zyx_1[ci] >= image.shape[ci]: zyx_1_fix[ci] = image.shape[ci] if zyx_2[ci] < 0: zyx_2_fix[ci] = 0 elif zyx_2[ci] >= image.shape[ci]: zyx_2_fix[ci] = image.shape[ci] img_crop = image[zyx_1_fix[0]:zyx_2_fix[0], zyx_1_fix[1]:zyx_2_fix[1], zyx_1_fix[2]:zyx_2_fix[2]] img_crop[img_crop < -1024] = -1024 #if img_crop.max() >= 600: # padding_value = 600 #elif img_crop.max() >= 0: # padding_value = img_crop.max() #else: # padding_value = -1024 padding_value = -1024 nodule_box = padding_value * np.ones( [box_size, box_size, box_size], int) nodule_box[zyx_1_fix[0] - zyx_1[0]:zyx_2_fix[0] - zyx_1[0], zyx_1_fix[1] - zyx_1[1]:zyx_2_fix[1] - zyx_1[1], zyx_1_fix[2] - zyx_1[2]:zyx_2_fix[2] - zyx_1[2]] = img_crop else: #nodule_box = np.zeros([box_size, box_size, box_size], np.int16) img_crop = image[zyx_1[0]:zyx_2[0], zyx_1[1]:zyx_2[1], zyx_1[2]:zyx_2[2]] # ---截取立方体 img_crop[img_crop < -1024] = -1024 # ---设置窗宽,小于-1024的体素值设置为-1024 nodule_box = img_crop[0:box_size, 0:box_size, 0:box_size] # ---将截取的立方体置于nodule_box self.save_annotations_nodule( nodule_box, patient_uid + "_" + annocol + "_ob") print("annotation sampling done") #get candidate annotation nodule candidate_coords = [] if candsample: segimage, segmask, flag = cd.segment_lung_mask(image) if segimage is not None: nodule_matrix, index = cd.candidate_detection( segimage, flag) cluster_labels = lc.seed_mask_cluster(nodule_matrix, cluster_size=1000) #cluster_labels = lc.seed_volume_cluster(image, segmask, eliminate_lower_size=-1) #segresult = lc.segment_color_vision(image, cluster_labels) #cv.view_CT(segresult) #lc.cluster_size_vision(cluster_labels) candidate_coords, _ = lc.cluster_centers(cluster_labels) #candidate_coords = lc.cluster_center_filter(image, candidate_coords) #the coordination order is [z,y,x] print("candidate number:%d" % (len(candidate_coords))) #volume_regioned = cv.view_coordinations(image, candidate_coords, window_size=10, reverse=False, slicewise=True, show=False) #mt.write_mhd_file(self.vision_path+"/"+patient_uid+"_candidate.mhd", volume_regioned, volume_regioned.shape[::-1]) for cc in range(len(candidate_coords)): candidate_center = candidate_coords[cc] invalid_loc = False if mt.coord_overflow(candidate_center - window_half, image.shape) or mt.coord_overflow( candidate_center + BOX_SIZE - window_half, image.shape): invalid_loc = True continue for index_search, v_center_search in v_centers: rpos = v_center_search - candidate_center if abs(rpos[0]) < window_half and abs( rpos[1] ) < window_half and abs( rpos[2] ) < window_half: #the negative sample is located in the positive location invalid_loc = True break if not invalid_loc: zyx_1 = candidate_center - window_half zyx_2 = candidate_center + BOX_SIZE - window_half nodule_box = np.zeros( [BOX_SIZE, BOX_SIZE, BOX_SIZE], np.int16) #---nodule_box_size = 45 img_crop = image[zyx_1[0]:zyx_2[0], zyx_1[1]:zyx_2[1], zyx_1[2]:zyx_2[2]] #---截取立方体 img_crop[img_crop < -1024] = -1024 #---设置窗宽,小于-1000的体素值设置为-1000 if img_crop.shape[0] != BOX_SIZE | img_crop.shape[ 1] != BOX_SIZE | img_crop.shape[2] != BOX_SIZE: print("error in resmapleing shape") try: nodule_box[ 0:BOX_SIZE, 0:BOX_SIZE, 0: BOX_SIZE] = img_crop # ---将截取的立方体置于nodule_box except: print("random error") continue #nodule_box[nodule_box == 0] = -1024#---将填充的0设置为-1000,可能有极少数的体素由0=>-1000,不过可以忽略不计 self.save_nonnodule(nodule_box, patient_uid + "_cc_" + str(cc)) print("candidate sampling done") print('Done for this patient!\n\n') print('Done for all!')
def annotations_crop( self, randsample=True, candsample=False, overbound=False, augment=False ): #the term 'augment' is invalid when 'overbound' is True if os.access(self.output_path, os.F_OK): shutil.rmtree(self.output_path) os.makedirs(self.output_path) os.mkdir(self.nodules_npy_path) #训练用正样本路径 os.mkdir(self.all_annotations_mhd_path) #检查用正样本路径 os.mkdir(self.nonnodule_npy_path) #训练用负样本路径 os.mkdir(self.no_annotation_mhd_path) #检查用负样本路径 if not os.access(self.vision_path, os.F_OK): os.makedirs(self.vision_path) for patient in enumerate(tqdm(self.ls_all_patients)): patient = patient[1] #patient = './LUNA16/subset9\\1.3.6.1.4.1.14519.5.2.1.6279.6001.114914167428485563471327801935.mhd' print(patient) # 检查这个病人有没有大于3mm的结节标注 if patient not in self.df_annotations.file.values: print('Patient ' + patient + 'Not exist!') continue patient_uid = mt.get_serie_uid(patient) patient_nodules = self.df_annotations[self.df_annotations.file == patient] full_image_info = sitk.ReadImage(patient) full_scan = sitk.GetArrayFromImage(full_image_info) origin = np.array( full_image_info.GetOrigin())[::-1] #---获取“体素空间”中结节中心的坐标 old_spacing = np.array(full_image_info.GetSpacing() )[::-1] #---该CT在“世界空间”中各个方向上相邻单位的体素的间距 image, new_spacing = mt.resample(full_scan, old_spacing) #---重采样 print('resample done') v_centers = [] center_coords = [] for index, nodule in patient_nodules.iterrows(): nodule_diameter = nodule.diameter_mm nodule_center = np.array( [nodule.coordZ, nodule.coordY, nodule.coordX]) #---获取“世界空间”中结节中心的坐标 v_center = np.rint( (nodule_center - origin) / new_spacing) #映射到“体素空间”中的坐标 v_center = np.array(v_center, dtype=int) v_centers.append([index, nodule_diameter, v_center]) center_coords.append(v_center) #volume_regioned = cvm.view_coordinations(image, center_coords, window_size=int(math.ceil(1.5*nodule_diameter)), reverse=False, slicewise=False, show=False) #np.save(self.vision_path+"/"+patient_uid+"_annotated.npy", volume_regioned) #---这一系列的if语句是根据“判断一个结节的癌性与否需要结合该结节周边位置的阴影和位置信息”而来,故每个结节都获取了比该结节尺寸略大的3D体素 #get annotations nodule window_half = int(BOX_SIZE / 2) if overbound: num_translations = 1 for index, nodule_diameter, v_center in v_centers: zyx_1 = v_center - BOX_SIZE # 注意是: Z, Y, X zyx_2 = v_center + BOX_SIZE if mt.coord_overflow(zyx_1, image.shape) or mt.coord_overflow( zyx_2, image.shape): continue nodule_box = np.zeros( [2 * BOX_SIZE, 2 * BOX_SIZE, 2 * BOX_SIZE], np.int16) # ---nodule_box_size = 45 img_crop = image[zyx_1[0]:zyx_2[0], zyx_1[1]:zyx_2[1], zyx_1[2]:zyx_2[2]] # ---截取立方体 img_crop[img_crop < -1024] = -1024 # ---设置窗宽,小于-1024的体素值设置为-1024 try: nodule_box = img_crop[ 0:2 * BOX_SIZE, 0:2 * BOX_SIZE, 0:2 * BOX_SIZE] # ---将截取的立方体置于nodule_box except: print("annotation error") continue #nodule_box[nodule_box == 0] = -1024 # ---将填充的0设置为-1000,可能有极少数的体素由0=>-1000,不过可以忽略不计 self.save_annotations_nodule( nodule_box, patient_uid + "_" + str(index) + "_ob") else: if not augment: scales = [1.0] translations = np.array([0, 0, 0]) else: scales = [0.8, 1.0, 1.25] #translations = np.array([[0,0,0],[0,0,1],[0,0,-1],[0,1,0],[0,-1,0],[1,0,0],[-1,0,0]], dtype=float) translations = np.array( [[0, 0, 0], [0, 0, 1], [0, 0, -1], [0, 1, 0], [0, math.sqrt(0.5), math.sqrt(0.5)], [0, math.sqrt(0.5), -math.sqrt(0.5)], [0, -1, 0], [0, -math.sqrt(0.5), math.sqrt(0.5)], [0, -math.sqrt(0.5), -math.sqrt(0.5)], [1, 0, 0], [math.sqrt(0.5), 0, math.sqrt(0.5)], [math.sqrt(0.5), 0, -math.sqrt(0.5)], [math.sqrt(0.5), math.sqrt(0.5), 0], [ math.sqrt(0.3333), math.sqrt(0.3333), math.sqrt(0.3333) ], [ math.sqrt(0.3333), math.sqrt(0.3333), -math.sqrt(0.3333) ], [math.sqrt(0.5), -math.sqrt(0.5), 0], [ math.sqrt(0.3333), -math.sqrt(0.3333), math.sqrt(0.3333) ], [ math.sqrt(0.3333), -math.sqrt(0.3333), -math.sqrt(0.3333) ], [-1, 0, 0], [-math.sqrt(0.5), 0, math.sqrt(0.5)], [-math.sqrt(0.5), 0, -math.sqrt(0.5)], [-math.sqrt(0.5), math.sqrt(0.5), 0], [ -math.sqrt(0.3333), math.sqrt(0.3333), math.sqrt(0.3333) ], [ -math.sqrt(0.3333), math.sqrt(0.3333), -math.sqrt(0.3333) ], [-math.sqrt(0.5), -math.sqrt(0.5), 0], [ -math.sqrt(0.3333), -math.sqrt(0.3333), math.sqrt(0.3333) ], [ -math.sqrt(0.3333), -math.sqrt(0.3333), -math.sqrt(0.3333) ]]) num_translations = 3 for index, nodule_diameter, v_center in v_centers: for s in range(len(scales)): rt = np.zeros(num_translations, dtype=int) rt[1:num_translations] = np.random.choice( range(1, len(translations)), num_translations - 1, False) rt = np.sort(rt) for t in range(rt.size): scale = scales[s] box_size = int(np.ceil(BOX_SIZE * scale)) window_size = int(box_size / 2) translation = np.array(nodule_diameter / 2 * translations[rt[t]] / new_spacing, dtype=int) tnz = translation.nonzero() if tnz[0].size == 0 and t != 0: continue zyx_1 = v_center + translation - window_size # 注意是: Z, Y, X zyx_2 = v_center + translation + box_size - window_size if mt.coord_overflow( zyx_1, image.shape) or mt.coord_overflow( zyx_2, image.shape): continue nodule_box = np.zeros( [BOX_SIZE, BOX_SIZE, BOX_SIZE], np.int16) # ---nodule_box_size = 45 img_crop = image[zyx_1[0]:zyx_2[0], zyx_1[1]:zyx_2[1], zyx_1[2]:zyx_2[2]] # ---截取立方体 img_crop[ img_crop < -1024] = -1024 # ---设置窗宽,小于-1024的体素值设置为-1024 if not augment or scale == 1.0: img_crop_rescaled = img_crop else: img_crop_rescaled, rescaled_spacing = mt.resample( img_crop, new_spacing, new_spacing * scale) try: padding_shape = ( img_crop_rescaled.shape - np.array( [BOX_SIZE, BOX_SIZE, BOX_SIZE])) / 2 nodule_box = img_crop_rescaled[ padding_shape[0]:padding_shape[0] + BOX_SIZE, padding_shape[1]:padding_shape[1] + BOX_SIZE, padding_shape[2]:padding_shape[2] + BOX_SIZE] # ---将截取的立方体置于nodule_box except: # f = open("log.txt", 'a') # traceback.print_exc(file=f) # f.flush() # f.close() print("annotation error") continue #nodule_box[nodule_box == 0] = -1024 # ---将填充的0设置为-1000,可能有极少数的体素由0=>-1000,不过可以忽略不计 self.save_annotations_nodule( nodule_box, patient_uid + "_" + str(index) + "_" + str(s * rt.size + t)) print("annotation sampling done") #get candidate annotation nodule candidate_coords = [] if candsample: segimage, segmask, flag = cd.segment_lung_mask(image) if segimage is not None: #nodule_matrix, index = cd.candidate_detection(segimage,flag) #cluster_labels = lc.seed_mask_cluster(nodule_matrix, cluster_size=1000) cluster_labels = lc.seed_volume_cluster( image, segmask, eliminate_lower_size=-1) segresult = lc.segment_color_vision(image, cluster_labels) cvm.view_CT(segresult) #lc.cluster_size_vision(cluster_labels) exit() candidate_coords, _ = lc.cluster_centers(cluster_labels) #candidate_coords = lc.cluster_center_filter(image, candidate_coords) #the coordination order is [z,y,x] print("candidate number:%d" % (len(candidate_coords))) #volume_regioned = cv.view_coordinations(image, candidate_coords, window_size=10, reverse=False, slicewise=True, show=False) #mt.write_mhd_file(self.vision_path+"/"+patient_uid+"_candidate.mhd", volume_regioned, volume_regioned.shape[::-1]) for cc in range(len(candidate_coords)): candidate_center = candidate_coords[cc] invalid_loc = False if mt.coord_overflow(candidate_center - window_half, image.shape) or mt.coord_overflow( candidate_center + BOX_SIZE - window_half, image.shape): invalid_loc = True continue for index_search, nodule_diameter_search, v_center_search in v_centers: rpos = v_center_search - candidate_center if abs(rpos[0]) < window_half and abs( rpos[1] ) < window_half and abs( rpos[2] ) < window_half: #the negative sample is located in the positive location invalid_loc = True break if not invalid_loc: zyx_1 = candidate_center - window_half zyx_2 = candidate_center + BOX_SIZE - window_half nodule_box = np.zeros( [BOX_SIZE, BOX_SIZE, BOX_SIZE], np.int16) #---nodule_box_size = 45 img_crop = image[zyx_1[0]:zyx_2[0], zyx_1[1]:zyx_2[1], zyx_1[2]:zyx_2[2]] #---截取立方体 img_crop[img_crop < -1024] = -1024 #---设置窗宽,小于-1000的体素值设置为-1000 if img_crop.shape[0] != BOX_SIZE | img_crop.shape[ 1] != BOX_SIZE | img_crop.shape[2] != BOX_SIZE: print("error in resmapleing shape") try: nodule_box[ 0:BOX_SIZE, 0:BOX_SIZE, 0: BOX_SIZE] = img_crop # ---将截取的立方体置于nodule_box except: print("random error") continue #nodule_box[nodule_box == 0] = -1024#---将填充的0设置为-1000,可能有极少数的体素由0=>-1000,不过可以忽略不计 self.save_nonnodule(nodule_box, patient_uid + "_cc_" + str(cc)) print("candidate sampling done") #get random annotation nodule if randsample: if overbound: augnum = 100 elif augment: augnum = len(scales) * num_translations else: augnum = 1 if augnum * len(v_centers) > len(candidate_coords): randnum = augnum * len(v_centers) - len(candidate_coords) else: randnum = len(candidate_coords) for rc in range( randnum ): #the random samples is one-to-one number of nodules #index, nodule_diameter, v_center = v_centers[rc] rand_center = np.array([0, 0, 0]) # 注意是: Z, Y, X invalid_loc = True candidate_overlap = True while invalid_loc: invalid_loc = False candidate_overlap = False for axis in range(rand_center.size): rand_center[axis] = np.random.randint( 0, image.shape[axis]) if mt.coord_overflow(rand_center - window_half, image.shape) or mt.coord_overflow( rand_center + BOX_SIZE - window_half, image.shape): invalid_loc = True continue if 'segmask' in dir() and not ( segmask is None ) and not segmask[rand_center[0], rand_center[1], rand_center[2]]: invalid_loc = True continue for index_search, nodule_diameter_search, v_center_search in v_centers: rpos = v_center_search - rand_center if abs(rpos[0]) < window_half and abs( rpos[1] ) < window_half and abs( rpos[2] ) < window_half: #the negative sample is located in the positive location invalid_loc = True break for candidate_coord in candidate_coords: rpos = candidate_coord - rand_center if abs(rpos[0]) < window_half and abs( rpos[1] ) < window_half and abs( rpos[2] ) < window_half: #the negative sample is located in the pre-extracted candidate locations candidate_overlap = True break if candidate_overlap: continue zyx_1 = rand_center - window_half zyx_2 = rand_center + BOX_SIZE - window_half nodule_box = np.zeros([BOX_SIZE, BOX_SIZE, BOX_SIZE], np.int16) #---nodule_box_size = 45 img_crop = image[zyx_1[0]:zyx_2[0], zyx_1[1]:zyx_2[1], zyx_1[2]:zyx_2[2]] #---截取立方体 img_crop[ img_crop < -1024] = -1024 #---设置窗宽,小于-1000的体素值设置为-1000 if img_crop.shape[0] != BOX_SIZE | img_crop.shape[ 1] != BOX_SIZE | img_crop.shape[2] != BOX_SIZE: print("error in resmapleing shape") try: nodule_box[ 0:BOX_SIZE, 0:BOX_SIZE, 0:BOX_SIZE] = img_crop # ---将截取的立方体置于nodule_box except: # f = open("log.txt", 'a') # traceback.print_exc(file=f) # f.flush() # f.close() print("candidate error") continue #nodule_box[nodule_box == 0] = -1024#---将填充的0设置为-1000,可能有极少数的体素由0=>-1000,不过可以忽略不计 self.save_nonnodule(nodule_box, patient_uid + "_rc_" + str(rc)) print("random sampling done") print('Done for this patient!\n\n') print('Done for all!')
def detection(**kwargs): opt.parse(kwargs) test_paths = ["/data/fyl/datasets/Tianchi_Lung_Disease/train"] #test_sample_filelist = "/data/fyl/models_pytorch/DensecropNet_detection_test_rfold1/filelist_val_fold0.log" #net_file = "/data/fyl/models_pytorch/DensecropNet_stripe_detection_rfold1/DensecropNet_stripe_detection_rfold1_epoch27" annotation_file = "/data/fyl/datasets/Tianchi_Lung_Disease/chestCT_round1_annotation.csv" #candidate_file = "/data/fyl/datasets/Tianchi_Lung_Disease/candidate.csv" evaluation_path = "./experiments_dt/evaluations_tianchild_densecropnet_nodule_rfold1" #evaluation_path = "experiments_dt/evaluations_test" #vision_path = evaluation_path result_file = evaluation_path + "/result.csv" hard_negatives_file = evaluation_path + "/hard_negatives.csv" region_size = opt.input_size batch_size = opt.batch_size label_dict = { 'noduleclass': 1, 'stripeclass': 5, 'arterioclass': 31, 'lymphnodecalclass': 32 } label = label_dict[opt.label_mode] use_gpu = opt.use_gpu net_file = opt.load_model_path if 'vision_path' in dir() and vision_path is not None and not os.access( vision_path, os.F_OK): os.makedirs(vision_path) #if os.access(evaluation_path, os.F_OK): shutil.rmtree(evaluation_path) if not os.access(evaluation_path, os.F_OK): os.makedirs(evaluation_path) if "test_paths" in dir(): all_patients = [] for path in test_paths: all_patients += glob(path + "/*.mhd") if len(all_patients) <= 0: print("No patient found") exit() else: print("No test data") exit() if hasattr(opt, 'filelists') and 'test' in opt.filelists.keys(): test_samples = bt.filelist_load(opt.filelists['test']) test_uids = [] for test_sample in test_samples: sample_uid = os.path.basename(test_sample).split('_')[0] if sample_uid not in test_uids: test_uids.append(sample_uid) pd.DataFrame(data=test_uids, columns=['series_uid' ]).to_csv(result_path + '/patients_uid.csv', index=False) #else: # for path in opt.filelists['test']: # test_samples = glob(path + '/*.mhd') #model = models.DensecropNet(input_size=region_size, drop_rate=0, growth_rate=64, num_blocks=4, num_fin_growth=3).eval() model = getattr(models, opt.model)(input_size=region_size, **opt.model_setup).eval() if net_file is not None: model.load(net_file) print('model loaded from %s' % (net_file)) shutil.copyfile(net_file, evaluation_path + '/' + net_file.split('/')[-1]) #model.eval() if use_gpu: model.cuda() start_time = time.time() #patient_evaluations = open(evaluation_path + "/patient_evaluations.log", "w") results = [] CPMs = [] CPMs2 = [] hard_negatives = [] test_patients = all_patients #random.shuffle(test_patients) bt.filelist_store(test_patients, evaluation_path + "/patientfilelist.log") for p in range(len(test_patients)): patient = test_patients[p] #patient = "./LUNA16/subset9/1.3.6.1.4.1.14519.5.2.1.6279.6001.212608679077007918190529579976.mhd" #patient = "./LUNA16/subset9/1.3.6.1.4.1.14519.5.2.1.6279.6001.102681962408431413578140925249.mhd" #patient = "./TIANCHI_examples/LKDS-00005.mhd" uid = mt.get_mhd_uid(patient) if 'test_uids' in dir() and uid not in test_uids: print("%d/%d patient %s not belong to test set" % (p + 1, len(test_patients), uid)) continue #if uid!='656867': # continue print('%d/%d processing patient:%s' % (p + 1, len(test_patients), uid)) full_image_info = sitk.ReadImage(patient) full_scan = sitk.GetArrayFromImage(full_image_info) origin = np.array(full_image_info.GetOrigin( ))[::-1] #the order of origin and old_spacing is initially [z,y,x] old_spacing = np.array(full_image_info.GetSpacing())[::-1] image, new_spacing = mt.resample(full_scan, old_spacing, np.array([1, 1, 1])) #image = np.load(patient) #new_spacing = np.array([1, 1, 1]) #origin = np.array([0, 0, 0]) print('Resample Done. time:{}s'.format(time.time() - start_time)) if 'annotation_file' in dir(): annotations = mt.get_challenge_annotations(uid, annotation_file, label) if len(annotations) == 0: print("%d/%d patient %s has no annotations, ignore it." % (p + 1, len(test_patients), uid)) #patient_evaluations.write('%d/%d patient %s has no annotations, ignore it\n' %(p+1, len(test_patients), uid)) continue #make a real lesion visualization if 'vision_path' in dir() and vision_path is not None: real_lesions = [] for annotation in annotations: #real_lesion = np.int_([abs(annotation[2]-origin[0])/new_spacing[0], abs(annotation[1]-origin[1])/new_spacing[1], abs(annotation[0]-origin[2])/new_spacing[2]]) real_lesion = mt.coord_conversion(annotation[:3][::-1], origin, old_spacing, full_scan.shape, image.shape, dir_array=True) real_lesions.append(real_lesion) annotation_vision = cvm.view_coordinates(image, real_lesions, window_size=10, reverse=False, slicewise=False, show=False) np.save(vision_path + "/" + uid + "_annotations.npy", annotation_vision) if 'candidate_file' in dir(): print('Detection with given candidates:{}'.format(candidate_file)) candidate_coords = nd.luna_candidate(image, uid, origin, new_spacing, candidate_file, lung_segment=True, vision_path=vision_path) if 'vision_path' in dir() and vision_path is not None: volume_candidate = cvm.view_coordinates(image, candidate_coords, window_size=10, reverse=False, slicewise=False, show=False) np.save(vision_path + "/" + uid + "_candidate.npy", volume_candidate) print('Candidate Done. time:{}s'.format(time.time() - start_time)) print('candidate number:%d' % (len(candidate_coords))) candidate_predictions = nd.precise_detection_pt( image, region_size, candidate_coords, model, None, batch_size, use_gpu=use_gpu, prediction_threshold=0.4) positive_predictions = candidate_predictions > 0 predicted_coords = np.delete( candidate_coords, np.logical_not(positive_predictions).nonzero()[0], axis=0) predictions = candidate_predictions[positive_predictions] lesion_center_predictions = nd.prediction_combine( predicted_coords, predictions) if 'vision_path' in dir() and vision_path is not None: volume_predicted = cvm.view_coordinates(image, predicted_coords, window_size=10, reverse=False, slicewise=False, show=False) np.save(vision_path + "/" + uid + "_predicted.npy", volume_predicted) lesions = [] for nc in range(len(lesion_center_predictions)): lesions.append(np.int_(lesion_center_predictions[nc][0:3])) volume_prediction = cvm.view_coordinates(image, lesions, window_size=10, reverse=False, slicewise=False, show=False) np.save(vision_path + "/" + uid + "_prediction.npy", volume_prediction) else: print('Detection with slic candidates') candidate_results = nd.slic_candidate(image, 30, focus_area='lung') if candidate_results is None: continue candidate_coords, candidate_labels, cluster_labels = candidate_results if 'vision_path' in dir() and vision_path is not None: np.save(vision_path + "/" + uid + "_segmask.npy", cluster_labels) #segresult = lc.segment_vision(image, cluster_labels) #np.save(vision_path + "/" + uid + "_segresult.npy", segresult) print('Candidate Done. time:{}s'.format(time.time() - start_time)) print('candidate number:%d' % (len(candidate_coords))) candidate_predictions = nd.precise_detection_pt( image, region_size, candidate_coords, model, None, batch_size, use_gpu=use_gpu, prediction_threshold=0.4) positive_predictions = candidate_predictions > 0 result_predictions, result_labels = nd.predictions_map_fast( cluster_labels, candidate_predictions[positive_predictions], candidate_labels[positive_predictions]) if 'vision_path' in dir() and vision_path is not None: np.save(vision_path + "/" + uid + "_detlabels.npy", result_labels) np.save(vision_path + "/" + uid + "_detpredictions.npy", result_predictions) #detresult = lc.segment_vision(image, result_labels) #np.save(vision_path+"/"+uid+"_detresult.npy", detresult) lesion_center_predictions = nd.prediction_centering_fast( result_predictions) #lesion_center_predictions, prediction_labels = nd.prediction_cluster(result_predictions) if 'vision_path' in dir() and vision_path is not None: lesions = [] for nc in range(len(lesion_center_predictions)): lesions.append(np.int_(lesion_center_predictions[nc][0:3])) volume_predicted = cvm.view_coordinates(result_predictions * 1000, lesions, window_size=10, reverse=False, slicewise=False, show=False) np.save(vision_path + "/" + uid + "_prediction.npy", volume_predicted) if 'prediction_labels' in dir(): prediction_cluster_vision = lc.segment_color_vision( prediction_labels) np.save( vision_path + "/" + uid + "_prediction_clusters.npy", prediction_cluster_vision) print('Detection Done. time:{}s'.format(time.time() - start_time)) ''' #randomly create a result for testing lesion_center_predictions = [] for nc in range(10): lesion_center_predictions.append([random.randint(0,image.shape[0]-1), random.randint(0,image.shape[1]-1), random.randint(0,image.shape[2]-1), random.random()]) ''' if len(lesion_center_predictions) < 1000: print('Nodule coordinations:') if len(lesion_center_predictions) <= 0: print('none') for nc in range(len(lesion_center_predictions)): print('{} {} {} {}'.format(lesion_center_predictions[nc][0], lesion_center_predictions[nc][1], lesion_center_predictions[nc][2], lesion_center_predictions[nc][3])) for nc in range(len(lesion_center_predictions)): #the output coordination order is [x,y,z], while the order for volume image should be [z,y,x] result = [uid] result.extend( mt.coord_conversion(lesion_center_predictions[nc][:3], origin, old_spacing, full_scan.shape, image.shape, dir_array=False)[::-1]) if label is not None: result.append(label) result.append(lesion_center_predictions[nc][3]) results.append(result) #results.append([uid, (lesion_center_predictions[nc][2]*new_spacing[2])+origin[2], (lesion_center_predictions[nc][1]*new_spacing[1])+origin[1], (lesion_center_predictions[nc][0]*new_spacing[0])+origin[0], lesion_center_predictions[nc][3]]) #if len(lesion_center_predictions)<1000: #print('{} {} {} {}' .format(lesion_center_predictions[nc][0], lesion_center_predictions[nc][1], lesion_center_predictions[nc][2], lesion_center_predictions[nc][3])) columns = ['seriesuid', 'coordX', 'coordY', 'coordZ', 'probability'] if label is not None: columns.insert(4, 'class') result_frame = pd.DataFrame(data=results, columns=columns) result_frame.to_csv(result_file, index=False, float_format='%.4f') np.save(evaluation_path + '/result.npy', np.array(results)) if 'annotation_file' in dir(): assessment = eva.detection_assessment(results, annotation_file, label=label) if assessment is None: print('assessment failed') #patient_evaluations.write('%d/%d patient %s assessment failed\n' %(p+1, len(test_patients), uid)) continue #num_scans, FPsperscan, sensitivities, CPMscore, FPsperscan2, sensitivities2, CPMscore2, lesions_detected = assessment num_scans = assessment['num_scans'] FPsperscan, sensitivities = assessment['FROC'] CPMscore = assessment['CPM'] prediction_order = assessment['prediction_order'] lesions_detected = assessment['detection_cites'] if len(FPsperscan) <= 0 or len(sensitivities) <= 0: print("No results to evaluate, continue") else: eva.evaluation_vision(CPMs, num_scans, FPsperscan, sensitivities, CPMscore, lesions_detected, output_path=evaluation_path) #patient_evaluations.write('%d/%d patient %s CPM score:%f\n' %(p+1, len(test_patients), uid, single_assessment[6])) print('Evaluation Done. time:{}s'.format(time.time() - start_time)) num_positive = (lesions_detected >= 0).nonzero()[0].size for ndi in range(len(lesions_detected)): if results[prediction_order[ndi]][-1] <= 0.5 or ( lesions_detected[:ndi] >= 0).nonzero()[0].size == num_positive: break if lesions_detected[ndi] == -1: hard_negatives.append(results[prediction_order[ndi]]) hard_negatives_frame = pd.DataFrame(data=hard_negatives, columns=columns) hard_negatives_frame.to_csv(hard_negatives_file, index=False, float_format='%.4f') print('Hard Negatives Extracted. time:{}s'.format(time.time() - start_time)) print('Overall Detection Done')
def detection_fusion( test_path=None, result_path="./experiments_dt/evaluations_tianchild_densecropnet_31,32", **kwargs): opt.parse(kwargs) if test_path is None: test_paths = ["/data/fyl/datasets/Tianchi_Lung_Disease/train"] else: test_paths = [test_path] #test_sample_filelist = "/data/fyl/models_pytorch/DensecropNet_detection_test_rfold1/filelist_val_fold0.log" net_files = [ "/data/fyl/models_pytorch/DensecropNet_arterio_detection_rfold1/DensecropNet_arterio_detection_rfold1_epoch2", "/data/fyl/models_pytorch/DensecropNet_lymphnodecal_detection_rfold1/DensecropNet_lymphnodecal_detection_rfold1_epoch2" ] annotation_file = "/data/fyl/datasets/Tianchi_Lung_Disease/chestCT_round1_annotation.csv" #candidate_file = "/data/fyl/datasets/Tianchi_Lung_Disease/candidate.csv" labels = [31, 32] #result_path = "./experiments_dt/evaluations_tianchild_densecropnet_fusion" #vision_path = result_path #result_file = result_path + "/result.csv" hard_negatives_file = result_path + "/hard_negatives.csv" region_size = opt.input_size batch_size = opt.batch_size use_gpu = opt.use_gpu if 'vision_path' in dir() and vision_path is not None and not os.access( vision_path, os.F_OK): os.makedirs(vision_path) #if os.access(result_path, os.F_OK): shutil.rmtree(result_path) if not os.access(result_path, os.F_OK): os.makedirs(result_path) if "test_paths" in dir(): all_patients = [] for path in test_paths: all_patients += glob(path + "/*.mhd") if len(all_patients) <= 0: print("No patient found") exit() else: print("No test data") exit() if hasattr(opt, 'filelists') and 'test' in opt.filelists.keys(): test_samples = bt.filelist_load(opt.filelists['test']) test_uids = [] for test_sample in test_samples: sample_uid = os.path.basename(test_sample).split('_')[0] if sample_uid not in test_uids: test_uids.append(sample_uid) pd.DataFrame(data=test_uids, columns=['series_uid' ]).to_csv(result_path + '/patients_uid.csv', index=False) #else: # for path in opt.filelists['test']: # test_samples = glob(path + '/*.mhd') #model = models.DensecropNet(input_size=region_size, drop_rate=0, growth_rate=64, num_blocks=4, num_fin_growth=3).eval() networks = [ getattr(models, opt.model)(input_size=region_size, **opt.model_setup).eval() for m in range(len(net_files)) ] for n in range(len(net_files)): networks[n].load(net_files[n]) print('model loaded from %s' % (net_files[n])) shutil.copyfile(net_files[n], result_path + '/' + net_files[n].split('/')[-1]) if use_gpu: networks[n].cuda() start_time = time.time() #patient_evaluations = open(result_path + "/patient_evaluations.log", "w") results = [] labeled_results = [[] for l in range(len(labels))] CPMs = [[] for l in range(len(labels))] #hard_negatives = [] test_patients = all_patients #random.shuffle(test_patients) bt.filelist_store(test_patients, result_path + "/patientfilelist.log") for p in range(len(test_patients)): patient = test_patients[p] uid = mt.get_mhd_uid(patient) if 'test_uids' in dir() and uid not in test_uids: print("%d/%d patient %s not belong to test set" % (p + 1, len(test_patients), uid)) continue print('%d/%d processing patient:%s' % (p + 1, len(test_patients), uid)) full_image_info = sitk.ReadImage(patient) full_scan = sitk.GetArrayFromImage(full_image_info) origin = np.array(full_image_info.GetOrigin( ))[::-1] #the order of origin and old_spacing is initially [z,y,x] old_spacing = np.array(full_image_info.GetSpacing())[::-1] image, new_spacing = mt.resample(full_scan, old_spacing, np.array([1, 1, 1])) #image = np.load(patient) #new_spacing = np.array([1, 1, 1]) #origin = np.array([0, 0, 0]) print('Resample Done. time:{}s'.format(time.time() - start_time)) candidate_results = nd.slic_candidate(image, 20, focus_area='body') if candidate_results is None: continue candidate_coords, candidate_labels, cluster_labels = candidate_results if 'vision_path' in dir() and vision_path is not None: np.save(vision_path + "/" + uid + "_segmask.npy", cluster_labels) #segresult = lc.segment_vision(image, cluster_labels) #np.save(vision_path + "/" + uid + "_segresult.npy", segresult) print('Candidate Done. time:{}s'.format(time.time() - start_time)) print('candidate number:%d' % (len(candidate_coords))) candidate_predictions = nd.precise_detection_pt( image, region_size, candidate_coords, networks, None, batch_size, use_gpu=use_gpu, prediction_threshold=0.4) labeled_predictions = [] for l in range(len(labels)): label = labels[l] print('label: %d' % (label)) evaluation_path = result_path + '/' + str(label) if not os.access(evaluation_path, os.F_OK): os.makedirs(evaluation_path) if 'annotation_file' in dir(): annotations = mt.get_challenge_annotations(uid, annotation_file, label=label) if len(annotations) == 0: print("%d/%d patient %s has no annotations, ignore it." % (p + 1, len(test_patients), uid)) #patient_evaluations.write('%d/%d patient %s has no annotations, ignore it\n' %(p+1, len(test_patients), uid)) continue #make a real lesion visualization if 'vision_path' in dir() and vision_path is not None: real_lesions = [] for annotation in annotations: #real_lesion = np.int_([abs(annotation[2]-origin[0])/new_spacing[0], abs(annotation[1]-origin[1])/new_spacing[1], abs(annotation[0]-origin[2])/new_spacing[2]]) real_lesion = mt.coord_conversion(annotation[:3][::-1], origin, old_spacing, full_scan.shape, image.shape, dir_array=True) real_lesions.append(real_lesion) annotation_vision = cvm.view_coordinates(image, real_lesions, window_size=10, reverse=False, slicewise=False, show=False) np.save(evaluation_path + "/" + uid + "_annotations.npy", annotation_vision) positive_predictions = candidate_predictions[l] > 0 result_predictions, result_labels = nd.predictions_map_fast( cluster_labels, candidate_predictions[l][positive_predictions], candidate_labels[positive_predictions]) labeled_predictions.append(result_predictions) if 'vision_path' in dir() and vision_path is not None: np.save(evaluation_path + "/" + uid + "_detlabels.npy", result_labels) np.save(evaluation_path + "/" + uid + "_detpredictions.npy", result_predictions) #detresult = lc.segment_vision(image, result_labels) #np.save(evaluation_path+"/"+uid+"_detresult.npy", detresult) lesion_center_predictions = nd.prediction_centering_fast( result_predictions) #lesion_center_predictions, prediction_labels = nd.prediction_cluster(result_predictions) if 'vision_path' in dir() and vision_path is not None: lesions = [] for nc in range(len(lesion_center_predictions)): lesions.append(np.int_(lesion_center_predictions[nc][0:3])) volume_predicted = cvm.view_coordinates(result_predictions * 1000, lesions, window_size=10, reverse=False, slicewise=False, show=False) np.save(evaluation_path + "/" + uid + "_prediction.npy", volume_predicted) if 'prediction_labels' in dir(): prediction_cluster_vision = lc.segment_color_vision( prediction_labels) np.save( evaluation_path + "/" + uid + "_prediction_clusters.npy", prediction_cluster_vision) print('Detection Done. time:{}s'.format(time.time() - start_time)) ''' #randomly create a result for testing lesion_center_predictions = [] for nc in range(10): lesion_center_predictions.append([random.randint(0,image.shape[0]-1), random.randint(0,image.shape[1]-1), random.randint(0,image.shape[2]-1), random.random()]) ''' for nc in range(len(lesion_center_predictions)): #the output coordination order is [x,y,z], while the order for volume image should be [z,y,x] result = [uid] result.extend( mt.coord_conversion(lesion_center_predictions[nc][:3], origin, old_spacing, full_scan.shape, image.shape, dir_array=False)[::-1]) if label is not None: result.append(label) result.append(lesion_center_predictions[nc][3]) #results.append(result) labeled_results[l].append(result) columns = [ 'seriesuid', 'coordX', 'coordY', 'coordZ', 'probability' ] if label is not None: columns.insert(4, 'class') result_frame = pd.DataFrame(data=labeled_results[l], columns=columns) result_frame.to_csv("{}/result_{}.csv".format( evaluation_path, label), index=False, float_format='%f') #np.save("{}/result_{}.npy"%(evaluation_path, label), np.array(results)) if 'annotation_file' in dir(): assessment = eva.detection_assessment(labeled_results[l], annotation_file, label=label) if assessment is None: print('assessment failed') #patient_evaluations.write('%d/%d patient %s assessment failed\n' %(p+1, len(test_patients), uid)) continue #num_scans, FPsperscan, sensitivities, CPMscore, FPsperscan2, sensitivities2, CPMscore2, lesions_detected = assessment num_scans = assessment['num_scans'] FPsperscan, sensitivities = assessment['FROC'] CPMscore = assessment['CPM'] prediction_order = assessment['prediction_order'] lesions_detected = assessment['detection_cites'] if len(FPsperscan) <= 0 or len(sensitivities) <= 0: print("No results to evaluate, continue") else: eva.evaluation_vision(CPMs[l], num_scans, FPsperscan, sensitivities, CPMscore, lesions_detected, output_path=evaluation_path) #patient_evaluations.write('%d/%d patient %s CPM score:%f\n' %(p+1, len(test_patients), uid, single_assessment[6])) print('Evaluation Done. time:{}s'.format(time.time() - start_time)) labeled_predictions = np.array(labeled_predictions) prediction_labels = np.argmax(labeled_predictions, axis=0) predictions_fusion = labeled_predictions.sum(axis=0) / 4.0 fused_center_predictions = nd.prediction_centering_fast( predictions_fusion) if 'vision_path' in dir() and vision_path is not None: np.save(vision_path + "/" + uid + "_classlabels.npy", prediction_labels) for lcp in range(len(fused_center_predictions)): #the output coordination order is [x,y,z], while the order for volume image should be [z,y,x] center = fused_center_predictions[lcp] result = [uid] result.extend( mt.coord_conversion(center[:3], origin, old_spacing, full_scan.shape, image.shape, dir_array=False)[::-1]) result.append(labels[prediction_labels[center[0], center[1], center[2]]]) result.append(center[3]) results.append(result) columns = ['seriesuid', 'coordX', 'coordY', 'coordZ', 'probability'] if label is not None: columns.insert(4, 'class') result_frame = pd.DataFrame(data=results, columns=columns) result_frame.to_csv(result_path + '/result.csv', index=False, float_format='%f') np.save(result_path + '/result.npy', np.array(results)) print('Overall Detection Done')
))[::-1] #the order of origin and old_spacing is initially [z,y,x] old_spacing = np.array(full_image_info.GetSpacing())[::-1] image, new_spacing = mt.resample(full_scan, old_spacing) #resample print('Resample Done') segmask = lps.segment_lung_mask(image) #segmask = segmask[195:200] #image = image[195:200] segimage = image.copy() segimage[segmask == 0] = 1024 #segment the lung at the image print('Lung Segmentation Done') #cluster_labels = lc.seed_segment(image, segmask, cluster_size=3000, view_result=True) nodule_matrix, cindex = cd.candidate_detection(segimage) cluster_labels = lc.seed_volume_cluster(nodule_matrix, cluster_size=30, result_vision=False) print('Clustering Done') cluster_labels_merged = lc.cluster_merge(segimage, cluster_labels) print('Clusters Merged') lc.view_segment(image, cluster_labels_merged) candidate_coords = lc.cluster_centers(cluster_labels) volume_candidated = cv.view_coordinations(image, candidate_coords, window_size=10, reverse=False, slicewise=True, show=True) mt.write_mhd_file(vision_path + "/" + uid + "_candidate.mhd", volume_candidated, volume_candidated.shape[::-1]) print('Candidate Done')
import numpy as np import pycuda.driver as pcd from toolbox import Lung_Cluster as lc if __name__ == "__main__": label_exam = np.load("detection_vision/slic_labels_filtered.npy") clcenters1, cllabels1 = lc.cluster_centers_fast(label_exam) print clcenters1 clcenters2, cllabels2 = lc.cluster_centers_gpu(label_exam) print clcenters2